In 2024 Imaging Data Commons team processed CT images available in the National Lung Screening Trial (NLST) collection using TotalSegmentator and pyradiomics to segment anatomical organs and extract basic first order (e.g., mean intensity) and shape features (e.g., volume), and released all of the extracted quantitative features publicly in the IDC TotalSegmentator-CT-Segmentations
analysis results collection [1,2]. A basic Looker dashboard for exploring this dataset is available here.
The motivation for that work was to utilize segmentation as a tool - as opposed as the objective of research endeavor - to enable downstream use of the data for exploration and hypothesis generation by the users who do not have expertise/time/resources to extract those quantitative features on their own.
To date, one study was published utilizing liver mean attenuation measurements for this dataset to explore a novel biomarker of liver disease [3]. We also investigated detection of outliers/erroneous segmentation results in this dataset [4].
In this project we aim to improve materials/resources that accompany the dataset to make it more accessible to the community, and explore research directions where the extracted quantitative features can be useful. AI will facilitate and enable this project at multiple levels: for segmenting the regions of interest, extraction of radiomics features, analyzing relevant research literature for identifying hypotheses of interest, and facilitating exploration of those hypothesis on the real data.
Individual images/segmentations can be loaded in Slicer using SlicerIDCExplorer extension.
A basic Looker dashboard for exploring this dataset is available here, screenshot above displays distribution of volume of the T8 vertebra.
Plot from the Krishnaswamy et al. 2024 study summarizing concordance between the trends observed for the volume measurements of the total vertebrae volume derived using TotalSegmentator on a cohort of ~20,000 subjects vs a published study Limthongkul et al. 2010 measuring vertebral body volume conducted on the data from 40 subjects.
[1] Thiriveedhi, V. K., Krishnaswamy, D., Clunie, D., Pieper, S., Kikinis, R. & Fedorov, A. Cloud-based large-scale curation of medical imaging data using AI segmentation. Research Square (2024). https://doi.org/10.21203/rs.3.rs-4351526/v1
[2] Thiriveedhi, V. K., Krishnaswamy, D., Clunie, D. & Fedorov, A. TotalSegmentator segmentations and radiomics features for NCI Imaging Data Commons CT images. (2024). https://doi.org/10.5281/zenodo.8347012
[3] Weiss, J., Bernatz, S., Johnson, J., Thiriveedhi, V., Mak, R. H., Fedorov, A., Lu, M. T. & Aerts, H. J. W. Opportunistic assessment of steatotic liver disease in lung cancer screening eligible individuals. J. Intern. Med. (2025). https://doi.org/10.1111/joim.20053
[4] Krishnaswamy, D., Thiriveedhi, V. K., Ciausu, C., Clunie, D., Pieper, S., Kikinis, R. & Fedorov, A. Rule-based outlier detection of AI-generated anatomy segmentations. arXiv [eess.IV] (2024). at http://arxiv.org/abs/2406.14486